PHYSICAL REVIEW SPECIAL TOPICS - ACCELERATORS AND BEAMS 8, 072401 (2005)

Adjustable, short focal length permanent-magnet quadrupole based electron beam

final focus system
J. K. Lim,* P. Frigola, G. Travish, and J. B. Rosenzweig
Department of Physics and Astronomy, University of California at Los Angeles, Los Angeles, California 90095, USA

S. G. Anderson, W. J. Brown, J. S. Jacob, C. L. Robbins, and A. M. Tremaine

Lawrence Livermore National Laboratory, Livermore, California 94550, USA
(Received 8 March 2005; published 15 July 2005)
Advanced high-brightness beam applications such as inverse-Compton scattering (ICS) depend on
achieving of ultrasmall spot sizes in high current beams. Modern injectors and compressors enable the
production of high-brightness beams having needed short bunch lengths and small emittances. Along with
these beam properties comes the need to produce tighter foci, using stronger, shorter focal length optics.
An approach to creating such strong focusing systems using high-field, small-bore permanent-magnet
quadrupoles (PMQs) is reported here. A final-focus system employing three PMQs, each composed of 16
neodymium iron boride sectors in a Halbach geometry has been installed in the PLEIADES ICS
experiment. The field gradient in these PMQs is 560 T=m, the highest ever reported in a magnetic optics
system. As the magnets are of a fixed field strength, the focusing system is tuned by adjusting the position
of the three magnets along the beam line axis, in analogy to familiar camera optics. This paper discusses
the details of the focusing system, simulation, design, fabrication, and experimental procedure in creating
ultrasmall beams at PLEIADES.

DOI: 10.1103/PhysRevSTAB.8.072401 PACS numbers: 41.85.Lc, 75.50.Ww, 41.50.+h

I. INTRODUCTION The difficulty in matching is due to two effects based in

In recent years, there has been a revolution in the design
and implementation of electron sources due to the intro-
duction of the rf photoinjector [1], which produces beams
that have very high brightness, defined as simultaneous
high peak current and low transverse emittance. The ap-
plications of high-brightness electron beams generated by
such devices has also continued apace in this time. These
applications, e.g., driving a plasma wakefield accelerator
(PWFA) [2,3] or providing an electron beam for an inverse-
Compton scattering (ICS) x-ray source [4,5], are strongly
dependent on production of very dense beams. It has been
widely noted that the scaling of such applications to higher
performance levels (higher field in the PWFA, brighter for
the ICS source) requires shorter bunch lengths z and
lower transverse emittances "x;y , a situation that has strong
p =p2 
the method of preparing the beam: emittance compensa-
tion and pulse compression. When one compresses the
electron beam, significant energy spread is introduced. In
focusing the beam into the application interaction region,
one must simultaneously avoid chromatic aberrations due
to energy spread, and emittance growth due to residual
beam space-charge effects. Assuming one uses relatively
weak normal-conducting electromagnetic quads, to obtain
a small focal spot, the beam must first be expanded for final
focusing. In such a case, the final spot size, in the absence
of an elaborate, long chromatically compensated final-
focus system, is limited by chromatic aberrations. These
issues are illustrated by the following relation governing
the ratio of final spot size 
to initial 0 [7],
v
u


























































u
t 1  0 =f2 

implications for future development of electron beam   p; (1)

p =p  2 p
sources themselves. These considerations have led to in-
creased concentration on photoinjector optimization [6]
and development of pulse compression systems. Much
less emphasis has been placed, however, on the need to
p . When the initial  function 0  20 =" is larger
scale the dimensions of the beam focal systems down with
other relevant length scales—z , "x;y . The characteristic
length scale of the focal system is the minimum
‘‘ function,’’ 
x , which is related to the transverse
p











beam size at focus by 
x  
x "x . In order to achieve a
p . In this condition, the demagnifica-
small transverse beam size at focus, a high-gradient quad-
p =p. With larger initial beam size, the chro-
rupole focusing array is required.
*Electronic address: jlim@physics.ucla.edu
0 2
1  0 =f 1  
where the final term obtained in the limit =f
p=
than the effective focal length f of the final-focus system,
strong compression of the beam size is possible, up until
the chromatic aberrations begin to dominate. There is a
minimum in the beam size, for a fixed focal length system,



 p=
when p0 =f
tion is 2
matic contribution to the emittance grows rapidly and
because of the demagnification limit, the final beam size
actually grows with larger initial beam size.

1098-4402=05=8(7)=072401(17) 072401-1  2005 The American Physical Society

J. K. LIM et al.
An example, which serves as the experimental context
for the remainder of this work, serves to illustrate the
implications of Eq. (1) on the choice of focusing scheme.
A UCLA/LLNL collaboration has operated an ICS source
experiment termed the picosecond laser-electron interac-
tion for the dynamic evaluation of structures (PLEIADES)
[8,9], in which a 10 –100 TW, sub-ps, 820 nm laser pulse is
collided with a velocity compressed, sub-ps 50 –75 MeV
electron beam. With the present generation of large aper-
ture quads that UCLA employs, which have a maximum
gradient of 15 T=m over 10 cm length, the minimum
, 45
,
value of the effective focal length of the final focusing
with respect to their transverse symmetry plane.
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
we discuss an alternative scheme that sidesteps these diffi-
culties, by tuning the strength of the final-focus array (a
class of quadrupole triplet) only through the longitudinal
(z) positioning of the individual fixed-strength PMQs. In
this way, the issue of tunability is separated from both that
of alignment and that of achieving high gradient. Indeed,
one may access extremely large field gradients by using an
optimum, static intramagnet configuration. Such a magnet
configuration was introduced by Halbach [16]. This mag-
net employs a pure permanent-magnet geometry consisting
of 16 sections, with magnetization oriented at 0
and 90
triplet f  50 cm, and thus optimum demagnification for This type of magnet has been studied in the context of the

p =p  0:5% is only possible for 0  100 m, or (with compact linear collider (CLIC) program [12,13], and has
"x  "n =  0:14 mm mrad) 0  3:2 mm. We thus an- been implemented at the Cornell Electron Storage Ring
ticipate a minimum achievable spot size [Eq. (1)] of 
 (CESR) in the CLEO detector mini- system [17].
32 m. This spot-size estimate, which has been verified in
simulation, is unacceptably large for the ICS source, which
ultimately demands a factor of 4 smaller spot size for
nonlinear ICS interaction studies, as well as for increased
x-ray photon production.
In addition, by allowing the beam to expand before final
focus, space-charge forces (for beams of energy
<100 MeV) will increase the emittance [6], giving an
even larger final spot size. This limitation also precludes
more elaborate approaches, based on sextupole correction,
to eliminate chromatic aberrations. At low energy there-
fore, transport and final focusing systems must be compact;
short focal lengths are needed.
From this introduction, it is clear that one must look into
the use of much shorter focal length lenses in order to
create minimum  functions small enough for many mod-
ern electron beam applications. One might consider super-
conducting quadrupoles or solenoids, but these devices are
neither inexpensive nor easy to build with the necessary
small (cm-scale) dimensions. On the other hand,
permanent-magnet-based quadrupoles (PMQs) can access
high-field gradients, and have been under study recently in
two arenas: linear collider (LC) main accelerator focus-
ing—to mitigate the cost of the needed power supplies—
and final-focus magnets in both circular accelerators [10]
and LCs [11–14].
In the LC main accelerator application, an effort has
been made to design a moderate strength, but tunable
magnet [11]. Much of the international linear collider
(ILC) oriented work has been concentrated on tunability,
while simultaneously holding the magnetic center stable,
as required for use in the beam-based alignment process. In
addition, both ILC and CESR groups have studied the
influence of their respective expected radiation environ-
ments on the magnetization of the permanent-magnet (PM)
material [15], and found the effects to be small.
On the other hand, the need for extremely large field
gradients makes the introduction of significant gradient
tuning into a PMQ very challenging. In the present work,
072401-2
The details of 3D field calculation in the Halbach sec-
tioned PMQ geometry are discussed in the following sec-
tion. This geometry is known to allow field gradients
approaching the ratio 2Br =ri , where Br is the remanent
field of the PM material, and ri is the pole inner radius. For
the material used in the present study, NdFeB, Br 
1:22 T, and thus a 2.5 mm inner radius may produce over
600 T=m, a field gradient that is appropriate for use in the
PLEIADES ICS experiment.
This can be seen from the following argument. Because
the focal length of the system as a whole should be less
than 10 cm, the individual PMQ length is chosen to be
1 cm. If one restricts, as is appropriate for a triplet con-
figuration, the focusing phase advance (for the initial
operating energy of 70 MeV) to at most one-half [kq lq 
B0 =BRlq  0:5] in a single 1 cm lens, then over a
600 T=m field gradient is deduced. We shall see below
that this estimate on appropriate field gradient indeed
yields effective guidelines for optimizing the beam optics
in the PLEIADES final-focus system.
To allow an appreciation of the scope of challenges
encountered in implementing such a tunable, ultrashort
focal length final focusing array, we must present work
on: magnet theory and design; beam dynamics simulation;
fabrication, alignment and testing; and experimental beam
measurements. In the following treatment, we therefore
employ the following structure: the basic theory and simu-
lation of a single PMQ magnet is discussed in Sec. II; the
simulation of a focusing lattice is examined in Sec. III;
Sec. IV is dedicated to the design and fabrication of the
final-focus system; the measurement of the completed
system is reviewed in Sec. V; and finally, implementation,
usage, and beam measurements are presented in Sec. VI.
II. MAGNET THEORY AND SIMULATIONS
In this section, a 2D analytical model originally ap-
proached by Halbach and a 3D simulation model are
presented to illustrate the dependence of the PMQ per-
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . .
formance on the geometric and magnetic design parame-
ters in the Halbach pure PM design.
We first discuss the two dimensional Halbach-type PMQ
model, reviewing results that were originally established
by Halbach. These results are then used to guide initial
magnetic and beam optics design decisions.
To estimate the achievable gradients one obtains with a
PMQ, we first assume a permanent-magnet geometry in
which the magnetization varies continuously with an azi-
muthal angle . The field gradient is given by
 
1 1
B0  2Br  : (2)
ri ro
With Br  1:22 T, ri  2:5 mm, and ro  7:5 mm, we
have B0  650 T=m. In practice, the field gradient is
most strongly controlled by the inner radius ri .
It is not possible to make a PMQ with a continuously
varying remanent field, but the use of uniformly magne-
tized segments can well approximate the pure quadrupole
case. This was recognized by Halbach, who studied the
approximation to the ideal geometry that can be achieved
by segmenting the magnet into M geometrically identical
pieces, which have magnetization rotated appropriately in
each segment, as shown in Fig. 1. The segmentation of the
magnet into M sections diminishes the field gradient pre-
dicted in Eq. (2) by a factor a2 . For the Halbach geometry
considered with 16 segments, the coefficient of quadrupole
term, a2 , is found to be 0.94 —the field gradient is dimin-
ished by only 6% from the ideal geometry.
The model analysis above omits the effect of finite
length of the PMQ. In order to address these effects, the
magnetostatic simulation code RADIA [18] was used in 3D
modeling and numerical analysis of a 16-piece Halbach
PMQ. The magnetostatic object can be formulated as a
FIG. 1. A cross section of the 16 segment PMQ and the
magnetization orientations shown as arrows.
072401-3
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
composition of a set of smaller geometric components
bounded by planar polygons which then are linked to
make up the desired magnet model. Furthermore, each
component can be segmented into smaller subdivisions
for improved accuracy.
The goal of our 3D simulation analysis is to obtain a set
of optimized PMQ dimensions which, consistent with a
given remanent field, yield an appropriately high-field
gradient. Our application demands that the overall dimen-
sion of the PMQ be consistent with the beam optics di-
mensions (few cm focal lengths for each lens), and that the
beam be passed through the small bore where the strong
field gradient is produced. The design analysis also must
factor in technical difficulties in the manufacturing pro-
cess. NdFeB, a rare-earth PM material, was chosen for our
simulations over other rare-earth types, e.g., SmCo, be-
cause of our demand for a higher remanent field range.
Each component wedge block is bounded by two paral-
lel trapezoidal planes separated by a length of lz and a
given remanent field of Br . By implementing a simple loop
algorithm routine in RADIA, 16 wedge blocks are sequen-
tially stacked side-by-side where the final model design, as
seen in Fig. 1, is created. The final optimum PMQ dimen-
sions giving compact geometry and desired field gradient
determined from the simulations are ri  2:5 mm, ro 
7:5 mm, and lz  10:0 mm. The field gradient that pre-
dicted from this parameter set is B0  573 T=m. Note that
the gradient found in RADIA simulation is lower than that
deduced from Eq. (2) mainly due 3D (edge) effects. The
field gradient and the magnetic field profiles plotted in the
bore region are shown in Fig. 2. The effective magnetic
length determined from the simulation is leff  10:4 mm
(see Fig. 2).
We now discuss two of the most relevant factors con-
tributing to changes in the performance of the NdFeB
based PMQ: relaxation in PMQ due to intramagnetic in-
teractions, and partial demagnetization due to temperature
variations. The results of the demagnetization studies in
RADIA are used to determine whether the coercivity and
remanence in the NdFeB material are adequate.
The close proximity and high remanent fields of the
segments can influence the overall magnetic characteristics
of the assembled PMQs through mutual magnetic interac-
tions. Relaxation modeling was performed by iterative
calculations of the field strength and the magnetization of
each wedge block under the influence of the other 15
wedge blocks, until the values reach the specified precision
in the stabilized magnetization.
With the NdFeB magnet parameters from the material
datasheet and the calculation precision set in RADIA,
the simulation showed a negligible relaxation in the
chosen geometry of the PMQ. However, a permanent-
magnet material with a high coercivity is still desirable
for its high resistance against opposing external fields
and its resistance to temperature variation induced
demagnetization.
J. K. LIM et al. Phys. Rev. ST Accel. Beams 8, 072401 (2005)

Understanding the effect of temperature variation on the

0.6 permanent magnet is important in the situation where an
accidental exposure of a NdFeB PMQ to a temperature
0.4 elevation can cause demagnetization as, for instance, dur-
ing a vacuum bakeout. Demagnetization effects in the
0.2 simulation studies are introduced through a second-order
By (Tesla)

temperature dependence in the magnetization and the co-

0.0 ercivity equations as [19]:
-0.2 MT  MT0 1  a1 T  T0   a2 T  T0 2 ;(3a)
-0.4
HcJ T  HcJ T0 1  b1 T  T0   b2 T  T0 2 ;(3b)
Radia simulation
Hall magnetic-sensor where T0 is a reference temperature, and coefficients ai and
-0.6
bi are obtained from the material datasheet.
-1.0 -0.5 0.0 0.5 1.0 The relaxation of a PMQ magnetization is performed for
5 different chosen temperatures: 0, 20, 60, 90, and 120 C.
x (mm) In Fig. 3, the irreversible loss of a PMQ magnetization is
600

550
B’ (T/m)

500

450
-3 -2 -1 0 1 2 3
y (mm)

Radia simulation
0.3
Hall magnetic-sensor

0.2
By (Tesla)

0.1

0.0

-10 -5 0 5 10

z (mm)
FIG. 2. (Color) 3D code RADIA PMQ field: a comparison of
PMQ field profile measurements using a Hall probe to RADIA
simulation results (top), the field gradient (middle), and the
effective magnetic length taken at r  2:0 mm (bottom).
FIG. 3. (Color) Magnetic field spatial dependence showing de-
crease in slope as temperature value increases (top); percentage
loss of field gradient as a function of temperature (bottom).

072401-4
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . . Phys. Rev. ST Accel. Beams 8, 072401 (2005)

700
shown as a function of a temperature change. This loss is
small under 100 C, but it increases dramatically above
120 C. This behavior is similar to that obtained from 600
simulations reported in Ref. [19].
Geometric imperfections in PMQ due to fabrication and
assembly errors, and their effect on magnet performance 500

B' (T/m)
are investigated here in simulation. We concentrated our
efforts in three causes of PMQ imperfections: angular
orientation errors in the magnetic easy axis; sector-to- 400 1%
sector geometric errors; and PMQ bore aperture radius 2%
errors. From simulations, approximate allowable toleran- 3%
300
ces in mechanical errors for the individual modules are
obtained. A complete error budget, including alignment 5%
errors, is discussed in later sections. 200
In the first imperfection category, the following set of -3 -2 -1 0 1 2 3

errors are introduced into the magnetization easy axes of y (mm)

randomly selected PMQ segments: 0:225 , 0:450 ,
0:675 , and 1:125 , corresponding to absolute relative FIG. 5. (Color) Field gradient dependence on magnet wedge
errors of 1%, 2%, 3%, and 5%, respectively. The magnetic shape errors. Each curve represents wedge shape (mechanical)
field gradient is solved and presented for each of the deviation from the ideal.
angular error cases as shown in Fig. 4. As the magnitude
of the angular error increases, the field profile suffers from
particular set of wedge angle errors introduced, the varying
a shift in the magnetic center and introduction of variation
gap is tapered to the left side of the PMQ bore. The field
in the field gradient (higher-order multipoles) in the bore
gradient profile shows a significant sloping effect and a
region.
reduction in amplitude when the errors exceed 2%.
For the second type of imperfection, the following errors
As we shall discuss below, one of the final steps in the
in wedge angles (in this particular exercise, side length
manufacturing process is boring out the PMQ hole to a
error is excluded, as we shall discuss in the magnet fabri-
specified value, in our case 2.5 mm. In doing so, it is
cation, that final boring out of inner pole surface yields a
prudent to understand the effect of an undercut or overcut.
highly uniform surface) are chosen as the simulation input
As shown in the theoretical analysis, a variation in the
parameters: 0:225 , 0:450 , 0:675 , and 1:125 ;
PMQ bore radius is inversely proportional to the field. In
again corresponding to relative errors of 1%, 2%, 3%, and
the final case of our imperfect PMQ model, the following
5%, respectively, as shown in Fig. 5. The error parameters
set of PMQ bore radii are considered: 2.45, 2.50, and
are assigned to an arbitrary number of wedges in different
2.55 mm [Fig. 6]. As expected, the field gradient predicted
positions which consequently create varying gaps. For the

800
600

700
500

600 400
B' (T/m)

B' (T/m)

300
500
1%
2% 200
2.55 mm
400
3% 100
2.50 mm
5% 2.45 mm
300 0

-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

y (mm) y (mm)
FIG. 4. (Color) Field gradient dependence on magnetization FIG. 6. (Color) Field dependence on magnet bore radius: r 
orientation errors. 2:45 mm (top), 2.50 mm (middle), and 2.55 mm (bottom).

072401-5
J. K. LIM et al. Phys. Rev. ST Accel. Beams 8, 072401 (2005)

in different radius parameters gives 591 and 556 T=m,

corresponding to an inner radius of 2.45 and 2.55 mm,
respectively. This variation is B0  18 T=m. In com- x
paring with the 2D Halbach PMQ model, the variation is
deduced from Eq. (2) to be j2Br ri =r2i j  19:5 T=m.
The three notable error studies that have been performed
y
provide the following tolerances which have been specified
in our magnet design: (1) magnetization easy-axis angular
orientation errors are allowed up to 2% before a notable L1 L2 L3
change develops between the magnetic center and the f1 f2 f3
mechanical center; (2) wedge shape (angle) error are al-
lowed again up to 2%; and, finally, (3) bore radius varia- FIG. 7. (Color) Standard triplet design for a symmetric focus.
tions of 0:05 mm from the nominal 2.5 mm radius
produce an acceptable field gradient variation of 3% M21 x0 , where
from the ideal B0 . In Sec. III, magnet errors that are
L2 L2 1 1L2 =f3
incorporated into the final-focus system have deleterious 1 1  f3 1  f3  L1  f3  f2 
effects on the beam dynamics. The simulation of such M21     :
f3 f2 f1
effects are discussed next.
(4)

III. BEAM DYNAMICS This expression can be simplified by requiring a round

This section begins with the presentation on the theory
of optimum triplet configuration based on thin-lens analy-
sis. Then first-order beam simulations using the envelope
code TRACE3D are employed to show the overall feasibility
of the tunable final-focus system. Finally, with second-
order beam simulations performed with the multiparticle
code ELEGANT, the effects of magnet errors and chromatic
aberration effects on the final beam parameters are
investigated.
A. Optimized triplet design
In this section the limitations on the final beam spot size
using quadrupole triplet system are discussed and the
optimum configuration of a simple triplet design is found.
These design considerations motivate the choices of PMQ
focusing system configuration as well as upstream match-
ing optics.
Consider here an idealized system composed of three
thin lenses separated by drift spaces and alternating be-
tween focusing and defocusing in the horizontal (x) di-
mension. An illustration of this model problem is shown in
Fig. 7. The assumed constraints for this design are that the
focus be symmetric, and that the incoming beam is sym-
metric and at a waist, x0  y0 and 0x0  0y0  0. The
usual matrix formalism may be employed to describe the
particle trajectories in the system. Using Mf for a focusing
thin lens (with a change in sign of f in the case of a
defocusing lens) and ML for a drift, the transport matrix
of the system is simply M  ML3 Mf3 ML2 Mf2 ML1 Mf1 .
For a given maximum obtainable focusing strength, or
equivalently minimum focal length, fmin , one seeks the
parameters f1 , f2 , f3 , L1 , L2 , L3 that maximize the con-
vergence angle, x0 at the focal point. Ignoring the small
initial angle (x00  0), the final angle is given by x0f 
beam spot at the focus. This forces the matrix element
M21 to be equal for x and y. Setting the right-hand side of
Eq. (4) equal to its counterpart in y leads to the relation:
f1 f2  f1 f3  f2 f3  L1 L2  0: (5)
Similarly, requiring x  y at the third lens by equating the
~ 11 elements of the x and y versions of the matrix M ~
M
ML2 Mf2 ML1 Mf1 gives the equation
 
L
f1  f2 1  1 : (6)
L2
The expected result shown by Eq. (6) is that the second lens
must be stronger than the first, provided L1 > 0.
Applying Eqs. (5) and (6) to Eq. (4) and solving for
variables f2 , L1 , and L2 leads to
1 L1 L2
M21   : (7)
f22 L1  L2
Now, if it is assumed that the second lens provides the
shortest focal length, f2  fmin , the minimum spot size is
achieved by maximizing the expression L1 L2 =L1  L2 ,
subject to the constraints given by Eqs. (5) and (6) and the
fact that f3  fmin . These three conditions lead to the
2
inequality, L1 L2  fmin , which is sufficient to find that
L1 L2 =L1  L2  is maximized when L1  L2  fmin .
With back substitution, the optimum triplet configuration
has the following parameters:
L1  L2  L3  f2  f3  fmin and f1  2fmin :
(8)
Note that if Eqs. (4)–(6) were solved for f3 and the
assumption f3  fmin is made, the maximum convergence
angle still occurs with the parameters of Eq. (8), showing
this (a 2f, f, f asymmetric triplet lens configuration)

072401-6
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . .
to be the optimal design. We have adopted this design
philosophy in our PMQ-based final-focus system.
B. Simulations using TRACE3D
We employ the quad triplet configuration suggested by
the thin-lens analysis of the previous section, compactly
indicated by F-DD-FF. The double strength (half-focal
length) lenses are created by concatenating two identical
magnets of the type discussed in Sec. II. The double-length
magnets are moved in tandem, to act as a movable single
lens. This asymmetric triplet is first studied using TRACE3D
[20], an envelope code based on a first-order matrix de-
scription of the transport. The optimum lattice configura-
tion was studied with this code to establish tuning
conditions and performance predictions across a wide
range of beam energies. Further, as is discussed below,
the code is employed as a tuning guide in the experimental
application of the focusing system.
TRACE3D studies demonstrated the focusability of a wide
range of beam energies by dynamically configuring the
drift spaces between PMQ lenses. Two examples of such
simulation studies are illustrated in Fig. 8 for two different
beam energy settings: 72 and 92 MeV. The following beam
parameters (informed by upstream measurements at
PLEIADES) served as input: rms emittances of "x 
7 mm mrad, "y  15 mm mrad and  functions of x 
2:0 m, y  0:9 m. TRACE3D results, shown in Fig. 8, il-
lustrate that the magnet spacings L1 , L2 , and L3 between
quadrupole magnets expand with increasing electron beam
energy, as expected, in order to increase the focal power of
the triplet as it is applied to a more rigid beam.
Lower beam energy operation, below 60 MeV, is not
possible in a system where 560 T=m PMQs are utilized. To
access this lower energy range one may employ a weaker
focusing lens configuration. The following system has
been recently implemented for lower energy (20 –
35 MeV) ICS experiment mode: weak focus—strong de-
focus—strong focus (WF-SD-SF). Here the weak PMQ is
created by boring out an existing PMQ to a larger inner
radius, which produces a measured field gradient of
FIG. 8. (Color) Beam energy: 72 MeV (top) and 92 MeV
(bottom).
072401-7
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
270 T=m. In addition to replacing the first PMQ with a
weaker PMQ, the two following lenses consist of a single
PMQ as opposed to a double PMQ in the original configu-
ration. TRACE3D simulation of this weaker focusing array
which show system tunability down to 30 MeV, as has been
recently verified experimentally.
Detailed multiparticle simulation studies, performed
with a sophisticated higher-order matrix calculation code,
ELEGANT, of effects arising from the presence of errors on
the performance of the PMQ final-focus system are pre-
sented in the following section.
C. Simulations using ELEGANT
There are several types of intratriplet alignment errors
which may have profound effects upon the performance of
the final-focus system. The two most serious are rotated
(skew) PMQs and transverse magnetic center offset error.
A third source of error not addressed in the previous
section is the effect of chromatic aberrations.
Understanding the effects of these errors requires use of
a more powerful tool for analysis, the higher-order enve-
lope and tracking code ELEGANT [21]. The effects of these
errors are examined here separately. Solutions for mitigat-
ing the effects of the mechanical errors in the final-focus
system’s overall design are discussed below which effec-
tively correct errors to within tolerances allowed.
A systematically rotated field in a quadrupole magnet is
expected when the 4-fold symmetry of the quadrupole is
broken by an error in manufacturing process, for instance
of uneven wedge sector shapes and incorrectly magnetized
blocks. In addition, errors in angular placement of the
PMQs with respect to the beam line may also produce a
skewed magnet.
In the simulation code, the skew rotation term is indi-
vidually introduced in each quadrupole. In our study, we
outline a tolerance envelope in rotation by applying a
systematic error. A clockwise rotation error is applied to
focusing quadrupoles and a counterclockwise rotation er-
ror applied to defocusing quadrupoles. Based on the sim-
plified model constructed, ELEGANT simulations are
generated for three different cases of magnet rotation angle
errors: 0, 0.01, and 0.02 radians. The normalized emittan-
ces in both transverse dimensions display growth propor-
tional to the rotation error, as shown in Fig. 9. The final rms
beam spot size grows from 20 to 34 m when a 0.01 rad
systematic rotation error was introduced. The beam spot
size produced of course grows progressively worse with
increasing rotation; in the case of 0.02 rad rotation error, a
200% growth in final beam spot size is found.
In actuality, all of the PMQs shared the same magnet
characteristics, as they are all created by slicing each
magnet from a long assembly. Thus, a rotation error due
to a segment or magnetization variability found in one
particular magnet is found also in the other magnets and
therefore an azimuthal rotation error is essentially elimi-
J. K. LIM et al.
40 εnx
εny
εn (mm mrad)
30
20
10
0.00 0.01 0.02
Rotation error (rad)
FIG. 9. (Color) Rotation error effect on normalized emittance
growth through an otherwise ideal triplet, as simulated by
ELEGANT.
nated from the focusing system. Any other rotation errors
p=p  0:5%. The
can be also introduced through incorrect mounting of the
lens assembly. Therefore, considerable emphasis was
placed on mechanical tolerances. In particular, magnet
holders for enclosing PMQs are specifically machined to
comply with less than 0.01 rad angular error.
As noted in the magnet simulations of magnet easy-axis
orientation and wedge errors, the magnetic center is shifted
from the nominal geometric center of PMQ due to the
cumulative effect of magnet errors. When such PMQs are
installed into the translation rail of the final-focus system,
magnet centers of the focusing and the defocusing PMQs
will be in offset positions due to 90
rotation relation. The
effect of quad misalignments is to produce steering; if the
vertical midplane of a quadrupole is displaced vertically by
y, it will introduce a dipole field
Bx  B0
y. Further- becomes the dominant factor contributing to the final beam
more, this steering produces momentum dispersion, which
also leads to degraded performance. Determining the ac-
ceptable tolerance in magnet displacement places demands
on the types of adjustments one needs in mechanical design
of magnet holders. In practice, with magnetic center mea-
surement data, corrections are applied through adjust-
ments, after centerline measurement of each magnet, of
the PMQ holder.
To simulate misalignments, the PMQs are simply speci-
fied to have transverse misalignments in ELEGANT. The
focusing PMQs are assigned a vertical displacement error,
while the defocusing PMQs the same horizontal displace-
ment. The following values in displacement error are used:
0, 0.1, 0.2, and 0.4 mm. The displacement error’s effect on
the horizontal normalized emittance is less evident when
p =p
compared to the vertical emittance as the beam encounters
a greater maximum misalignment in this direction, as
shown in Fig. 10. The emittance growth due to misalign-
072401-8
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
18
εnx
17
16 εny
15
14
εn (mm mrad)
13
12
11
10
9
8
7
6
0.0 0.2 0.4
Displacement error (mm)
FIG. 10. (Color) Effect of magnet-center errors on emittance.
ment is less severe than that due to the skew rotation error,
because it is driven mainly by momentum dispersion,
which is a small effect for the design 
vertical emittance growth is near negligible up to 0.1 mm
misalignment, growing to 8% at 0.4 mm displacement from
the centers.
As noted in the introduction, the use of short focal length
magnets in the final-focus system is in fact motivated by
the need to evade chromatic effects on our ICS x-ray
experiment at PLEIADES. The focusing error from chro-
matic aberration causes a severe case of a beam demagni-
fication, which creates a low dense electron beam at the
interaction, and hence low ICS x rays are generated. The
effects of beam momentum spread effect on the final
focused beam spot have been simulated with ELEGANT,
and are shown in Fig. 11.
As mentioned in the introduction, chromatic aberration
size as the input beam size 0 is expanded. Since Eq. (1)
was derived as a generalization to a thick triplet lens, as
opposed to the original first-order, thin-lens treatment
found in Ref. [7], a simulation study of its validity was
performed using ELEGANT. The results of this analysis are
plotted in Fig. 12. In this study, we used the original design
emittance of the PLEIADES injector, so the overall scale
of the achievable spot sizes is smaller.
As can be seen in Fig. 12, chromatic aberration domi-
nates the initial beam emittance when the beam at final-
focus entrance 0 exceeds the optimum size,
s


















"x0
0 jopt  f: (9)

The corresponding minimum final beam spot size obtained
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . . Phys. Rev. ST Accel. Beams 8, 072401 (2005)

FIG. 11. (Color) ELEGANT simulation of chromatic effects: x and y phase space plots at the focal point; 
p=p  0 case (1st and 2nd)
and 
p=p  0:5%, showing chromatic aberrations (3rd and 4th).

18 from an engineering perspective. Assembling highly mag-

netized, small trapezoidal segments into a precisely
16 aligned unit with a small bore poses many mechanical
challenges. In addition, fabrication of multiple PMQs
14 with both identical magnetic and geometric characteristics
presents further difficulties.
( m)

12
Nevertheless, a simplified fabrication procedure for pro-
10
ducing a high quality, and ultrahigh gradient PMQ may be
f

given as follows. A long trapezoidal wedge which is pre-

8 magnetized to a Br  1:22 T is machined to a specified
design shape from a larger magnet block through a high
6 precision cutting process, wire electrical discharge ma-
chining (EDM). The 16 magnetized wedges are then
4
0 50 100 150 200 250 300
coupled together carefully into an equally long aluminum
keeper tube. In addition to magnet wedges being held
0
(m) inside the keeper, a thin layer of cohesion material is
applied, strongly joining neighboring magnets. The tube
FIG. 12. (Color) Study of chromatic aberration effects in keeper and a bonding glue together ensure structural stabil-
PLEIADES PMQ final-focus system, using ELEGANT simulation
ity of the PMQ under extreme magnetic repulsions. The
(dots), with theoretical prediction (line).
long assembly is then cut into six identical PMQs, with a
mechanical length of 10 mm, to ensure consistent perform-
in the presence of the chromatic aberrations is ance between all PMQs. In particular, construction of a
s















single, long magnet assembly assured that the relative


p






azimuthal (roll) errors of individual units may be essen-
 jmin  2f "x0 p : (10)
p tially eliminated. Finally, the inner diameter of bore is
slowly ground to the specified value of 5 mm with a
The simulations shown in Fig. 12 assume a round input
diamond cutter. The PMQ fabrication procedure is shown
beam with initial  functions in the range 2 m   
schematically in Fig. 13.
250 m, 0.6% fractional rms momentum spread and 3:6 
The manufacturing process required uniform magnet
102 mm mrad rms geometric emittances. The ELEGANT
material properties because no sorting process was relied
results are plotted against the prediction of Eq. (1). In this
upon when selecting for the individual magnet wedges.
way, the effective focal length of the final-focus system in
the x dimension was determined to be feff  6:2 cm. The
predicted optimum input  function for the minimum final
beam focal spot size of 5 m is 10 m for this case, which
has a relatively small emittance. For the actual experimen-
tal parameters with degraded emittance, the initial opti-
mum  is smaller.

IV. SYSTEM DESIGN AND PROPERTIES

A. Magnet design and fabrication
Fabrication of an ultrahigh field gradient Halbach-type
PMQ with good field quality has proven to be challenging FIG. 13. An illustration of the PMQ manufacturing process.

072401-9
J. K. LIM et al.
Instead, the long magnet wedges were all machined from a
single premagnetized block, with the field orientation as-
sured by the cut angle rather than a post-machining mag-
netization. This recipe transferred the possible sources of
error and precision to the machining process, and spared us
from having to measure very small magnets and sort
hundreds of wedges.
For the application at hand, a particular grade of neo-
dymium iron boride (NdFeB) was selected based on vari-
ous parameters, availability, and homogeneity. Key
parameters of the NdFeB (grade 35SH) used in the
PMQs are listed in Ref. [22].
B. Mover system
Before implementing the PMQ system on PLEIADES,
general problems associated with the application of this
focusing system in the experimental environment had to be
solved. Addressing the mechanical tolerances and space
constraints of the PMQ system, while maintaining an
environment consistent with ultrahigh vacuum (UHV)
and magnetic fields, required careful integration.
Precision machining of all parts eliminated the need for
adjustable alignment in the movers. Computer controlled
custom vacuum linear motion feedthroughs were utilized
to control each quadrupole’s longitudinal position, which
was precisely constrained by a rail system. Motion of the
PMQs was set by use of stepper motors and LabVIEW
control software. The entire magnet axis of the system was
aligned by use of a pulsed-wire system and optical tech-
niques. Other features of the assembly include a ceramic
ablation disc at the front of the assembly to protect the
magnets from misaligned high-intensity laser damage
(during the ICS experiment), and a scintillating disc at
the (beam entrance) side of the assembly, for locating
mis-steered electron beams.
From a mechanical design approach, the system had to
meet the following requirements: independent (preinstal-
lation) correction of magnetic center to mechanical center
errors for each PMQ assembly, independent position con-
trol of PMQs to within 100 m in the beam line direction,
magnetic center-to-center transverse placement with
10 m precision for each PMQ along the range of motion,
insertion of assembly into a standard 400 inner diameter
(ID) vacuum chamber, and compatibility with vacuum at a
nominal pressure of <106 Torr.
A mechanically simple design was chosen to comply
with these stringent requirements. A precision-machined
stainless steel Conflat flange was fabricated to serve as a
reference monument, mechanical support, and vacuum
seal. Three precision ground rods were fitted into the flange
by means of expansion and shrinkage fits.
Pairs of magnets were supported by precision-machined
aluminum disks, each with three linear bearings press fit
into the outer diameter. The disks and bearings move along
the rods, and are controlled by vacuum linear motion
072401-10
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
FIG. 14. (Color) Rendered CAD drawing of the final-focus
assembly.
feedthroughs. The entire mechanical assembly is supported
by the main flange using a massive precision ground stand,
which affords five-axis alignment, as shown in Fig. 14.
An initial set of disks were machined to place each PMQ
into the mechanical center of the assembly (  0:000500 ).
Using optical techniques along with a pulsed-wire system,
the magnetic center of each PMQ was found relative to the
assembled mechanical centers. New discs, if needed, were
then machined with appropriate offsets to eliminate these
errors. Fiducials were placed inside the front and back
PMQ for optical alignment.
V. PMQ SYSTEM MEASUREMENTS
Various measurements on the PMQs were performed
using pulsed wire and Hall probe. In comparison with the
3D magnetic simulations and tolerance studies, the quad-
rupole magnet properties indeed agree well with predicted
values. More details of measurement techniques are dis-
cussed in the following sections.
Measurements of the PMQs were made with the pulsed-
wire technique [23]. The test setup included a single meter-
long, 50 m diameter BeCu wire which was stretched out
through the PMQ magnetic field region and anchored at
end mounts. One end was fixed, while the other end had a
pulley over which the wire was tensioned by a hanging
weight. A pulse of 20 –50 V, with a duration of a few msec
to sec, and a repetition rate of 1 Hz was produced from a
high voltage pulser triggered by a function generator.
When the current pulse is passed along the wire through
the PMQ magnetic region, it interacts with the PMQ B field
and deflects the wire. A detector consisting of a photodiode
sensor and a diode laser was employed to quantify the
deflection. Two sets of laser-sensors were used in order
to measure both transverse deflections simultaneously.
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . . Phys. Rev. ST Accel. Beams 8, 072401 (2005)

The pulsed-wire technique was used to detect nonline-

arity in the B-field dependence on transverse displacement. 100 Pulsed-wire scan data

Pulsed wire signal (mV)

The BeCu wire was translated horizontally through the
magnetic center from 2:0 to 2:0 mm. The deflection
amplitude gives a measure of the (longitudinally) inte-
grated vertical component of the B field. According to
the data reported in Fig. 16, the transverse B field shows
no discernible higher multipole content; the dependence of
By on x is linear to within measurement error over the scan
interval.
The PMQ magnetic center to mechanical center align-
ment has also been studied using the pulsed-wire tech-
nique. A PMQ mounted on 2 m stages, with the wire
initially located at the magnet mechanical center, is con-
tinuously adjusted in transverse position until the magnetic
center is found by observing that the wire deflection signal
is reduced to near zero (see Fig. 15). From this measure-
ment, it was determined that the magnetic center was
aligned to the geometric center to within the instrument
resolution of 25 m.
The alignment of the final mover system was observed
with a theodolite for horizontal motion and an autolevel for
vertical motion. The alignment in the horizontal direction
was good to within 25 m. The autolevel indicated notable
sagging in the vertical direction, however. The sagging was
compensated by a set screw located on the bottom of the
ring piece which ties together the free ends of the rails.
Field gradient measurements of the PMQ were at-
tempted using the pulsed-wire technique. However, the
wavelength dependence of the wire response prevented
calibration of the 1 cm long magnet measurements, as
this was attempted using the known fields in a much longer
(10 cm) electromagnetic quadrupole. As an alternative
option for quantitative measurements, a miniature mag-
netic Hall probe was employed to determine the field
gradient and the effective magnetic length of the PMQ.
Linear-fit
2
50 R =0.9998
0
-50
-100
-2 -1 0 1 2
x (mm)
FIG. 16. (Color) Field linearity scan over accessible PMQ aper-
ture from pulsed-wire measurement.
With the PMQ mounted on the same translation microme-
ter stages, the probe scans horizontally reading By . The
field gradient obtained from this measurement in the center
of the PMQ bore region was measured to be 560 T=m (see
Fig. 2) using this method. We note that a Hall probe is
limited in spatial resolution, which in turn limits the pos-
sible characterization of nonlinear field components in a
quadrupole magnet.
An effective magnetic length was also determined using
a Hall probe by reading the value of By at each point along
a longitudinal line located 0.5 mm diagonally from the
magnet axial axis. The measured effective magnetic length
of 10.4 mm, as shown in Fig. 2, shows a good agreement
with the RADIA simulation. While the entire range of the
magnet is not accessible to measurement, the reproduction
of the RADIA results allows us to deduce the effective
magnet lengths, which are employed as input in the beam
dynamics simulations, to be essentially those given by
RADIA.

VI. BEAM-BASED MEASUREMENTS

A. The PLEIADES experiment
The PMQ system was installed in the PLEIADES ICS
experiment, in which a relativistic (25–80 MeV), high-
brightness electron beam is collided with a 50 fs, up to 1 J,
820 nm laser to produce short pulse (0.3–5 psec), high
peak brightness, hard x rays, tunable in energy from 40 –
140 keV.
The number of x-ray photons, Nx , produced by the
interaction of the laser and electron beams is determined
by the intensities of the beams as they overlap,
Z
FIG. 15. (Color) Single PMQ module on the pulsed-wire mea- ~
Nx  T c1  v~  k nL x;
~ tne x;
~ tdVdt; (11)
surement system. V;t

072401-11
J. K. LIM et al.
where nL x;~ t and ne x;
~ t are the laser photon and electron
beam densities, respectively, T is the total Thomson cross
section, v~ is the electron beam velocity, and k~ is the laser
propagation wave number. For the PLEIADES experiment,
a 180
scattering geometry is used. The integrated x-ray the high energy electron beam line was redesigned both to
flux is found, assuming Gaussian spatial and temporal
profiles for both beams, and that the laser and electron
beam pulse durations are less than the Rayleigh range and
 function of the laser and electron foci, respectively, to be
Ne NL T
Nx  : (12)
2*2e  2L 
Here e and L are the electron and laser beam spot sizes
at the interaction point, and Ne and NL are the total
numbers of electrons and laser photons in the pulses.
Note that PLEIADES experimental design parameters in-
clude a bunch charge approaching 1 nC (Ne  6  109 )
and a laser photon energy approaching 1 J at a wavelength
around 1 m, resulting in NL  5  1018 . Assuming the
incident laser and electron beams are matched such that
e  L , then to obtain a flux of x rays suitable for a single
shot diffraction experiment (about 108 ), an rms interaction
spot size in the tens of microns is required.
Further, since initial x-ray data were obtained with L <
e , the PLEIADES ICS experiment benefits significantly
from decreasing the electron beam size at the interaction
point and is therefore an excellent location to initially
implement the PMQ final-focus system.
B. Installation at PLEIADES
The PLEIADES electron beam is generated using a
BNL/SLAC/UCLA style 1.6 cell photocathode rf gun
[24]. Following the gun, the beam is accelerated to energies
of 20 –100 MeV using as many as four S-band 2.5 m
traveling wave sections. For the measurements described
below, the gun produced 200 –300 pC bunches with 3 ps
Compton scattering
interaction point
CCD for x-ray
detection
FALCON laser
820 nm, 1 J, 50 fsec
PMQ final focus system
FIG. 17. (Color) Schematic drawing of the PLEIADES interaction beam line.
072401-12
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
rms duration, while the final beam energy was varied
between 50 to 70 MeV, and the typical normalized, rms
emittance was around 10 mm mrad.
To install the PMQs as the e-beam final-focus system,
provide approximate beam matching into the PMQs and to
retain diagnostic capabilities. A schematic of this beam
line is shown in Fig. 17.
The laser-electron IP (interaction point) is located at the
center of a 6-inch cube vacuum chamber. As described
above, the PMQ mover assembly is mounted directly into
the interaction chamber, allowing positioning of the mag-
nets at a range of distances directly upstream of the IP (see
Fig. 18). The minimum distance of the final PMQ to the IP
is determined by the width of a polished aluminum cube
positioned at the IP. This cube functions as the spatial and
temporal overlap diagnostic by directing the laser light and
the optical transition radiation (OTR) produced by the
electron beam into a CCD camera and a streak camera.
In addition to enabling the overlap of the two beams, the
CCD images naturally allow optimization of the focused
spot size. This cube is extracted from the beam path for x-
ray production but cannot be fully removed from the PMQ
path. As a result, the minimum distance from the edge of
the final PMQ to the IP is 5 mm.
After the IP, a round pole dipole magnet is used to bend
the electron beam onto a trajectory which separates it from
the x-ray pulse, allowing the beam to be dumped into a
shielded Faraday cup, minimizing the noise reaching the x-
ray detectors. Note that a spectrometer magnet was also
located just upstream of the final matching quads to mea-
sure the energy and energy spread of the beam, which are
essential steps in tuning the linac.
C. Tuning the final-focus optics
In accordance with the quad triplet model used to design
the PMQ final-focus system, the beam must be manipu-
Dipole/Spectrometer
Steering magnets (energy spread)
Quadrupoles
(emittance and beam matching)
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . . Phys. Rev. ST Accel. Beams 8, 072401 (2005)

FIG. 18. (Color) The PMQ final-focus assembly after installation at PLEIADES. The vacuum cube enclosing the IP is shown
on the left.

lated upstream of the PMQs to produce a moderately
focused waist at the entrance to the first PMQ. The first
step in this process is the determination of the rms Twiss
parameters of the beam exiting the accelerator, which is
done using the quadrupole scan emittance measurement
technique. In the quad scan measurement, the current in the
first electromagnetic (EM) quad in Fig. 17 is varied while
the rms beam size is measured after a known drift length.
Taking into account the finite length of the quadrupole
magnet (using a ‘‘thick lens’’ analysis) the final beam
size in each transverse direction, f , follows the well-
known relation [25]
p



p



p




2f  cos K lq   K lq sin K lq 2 11;i
p



p




p




sin K lq 
2 p



 ld cos K lq  cos K lq 
K entrance is a compromise between emittance and aberra-
p



p




p



tion considerations, and will be discussed further below.
sin K lq 
 K ld sin K lq  12;i  p




K settings for the upstream optics, measuring the beam size
p



2
 ld cos K lq  22;i ; (13)
where lq;d are the EM quad and drift lengths, respectively,
K is the focusing strength of the quad given by its magnetic
field gradient and the beam momentum, K  eB0 =p, and
the initial Twiss parameters are i  11;i =", -i 
12;i =", and i  22;i =", where " is the rms emittance.
Because quad scans must be performed as part of the
routine operating procedure, the process was automated
using LabVIEW based control system software. The im-
mediate determination of the emittance provides the linac
operator with feedback on the quality of the linac tune, as
well as giving the Twiss parameters needed for matching
the beam into the PMQs.
Once the Twiss parameters have been determined, they
are used as input to a TRACE3D model of the beam line.
Given the beam parameters at the first element of the
model, and those desired at some latter point, TRACE3D
can perform a matching function, in which a user defined
set of model parameters is adjusted in an algorithm to
produce the desired final beam. In this case, we adjust
the strengths of EM quad doublet and triplet to produce a
beam waist (-  0) in both transverse dimensions at the
entrance to the PMQs with an rms size around 350 m. An
example of the output of this matching process is shown in
Fig. 19. The choice of rms beam size at the final-focus
Once the TRACE3D model has determined the proper
at two pop-in beam profile monitors and comparing the
results with the model predictions allows a partial check of
the model. With good agreement found between model and
measurement, TRACE3D is used again to determine the
proper spacing of the PMQs. The result of this matching
process is shown in Fig. 20. Here the spacing between
3 PMQ holders is adjusted to produce the smallest
 function possible at the IP.
The linac operator uses the TRACE3D values of upstream
quad strengths and PMQ positions as a starting point. One
important practical consideration the operator must take

072401-13
J. K. LIM et al.
FIG. 19. (Color) TRACE3D (screen shot) model of the beam line used to compute the focusing strength of the upstream EM quads for
proper matching into the PMQs.
into account is steering the beam onto the axis defined by
the PMQ mover system. This is needed both to minimize
the effects of chromatic and spherical aberrations in the
PMQs and to precisely define the direction of the ICS
x rays generated. There are two steering magnets located
between the EM quad triplet and the PMQs that are used
for this job. Steering is also made easier by imaging the
beam on a plastic scintillator mounted on the upstream face
of the first PMQ holder. This scintillator has a 5 mm
diameter hole that is mounted concentric to the PMQ
aperture. With these tools, proper beam steering is accom-
plished using the two steering magnets to iteratively move
the beam to the center of the scintillator hole and observe
the motion (or lack of motion) of the beam at the IP while
moving the final PMQ holder.
The process described above has been used to focus the
electron beam at the PLEIADES IP using the PMQ system.
The spot size was measured using OTR generated by the
beam colliding with the interaction cube diagnostic as
072401-14
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
described above. This light is collected by a CCD cam-
era/lens system and digitized using a computer frame
grabber. A typical and highly repeatable beam spot is
shown in Fig. 21. In this case we focused a 70 MeV,
200 pC beam to less than 20 m rms size in both trans-
verse dimensions. This is a significant improvement over
the previous final-focus system, which achieved spot sizes
of roughly 50 m, rms.
D. Emittance measurements with the PMQs
A variation of the quad scan technique for emittance
measurement is the so-called three screen or moving
screen measurement, in which there is a focusing lens
followed by a drift space and measurement screen.
However, in this case the beam size is measured while
varying the drift length instead of the lens strength. In the
simplified thin-lens analysis (where the focal length is f 
1=Klq ) the equation analogous to Eq. (13) is
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . . Phys. Rev. ST Accel. Beams 8, 072401 (2005)

FIG. 20. (Color) Typical TRACE3D output (screen shot) specifying the spacing between PMQs and their position with respect to the
beam focus.

 2 
 z   "0 1  zz
2



0
 2res ; (14)

where 
is the minimum  function at the waist, "0 is the

FIG. 21. (Color) False color CCD image of the PMQ focused
electron beam. The rms beam size is 20 m in both x and y.
beam emittance, and res is the image resolution obtained
from the limitations of optical transport and camera per-
formance. In the free drift, the square of the rms beam spot
size expands quadratically away from the beam waist
position, z0 .
In the case of the PMQ final-focus system, the focal
length, f, is fixed and the drift length, z, is changed most
easily by moving all three of the PMQ holders in unison.
This is not technically a correct application of Eq. (14),
since by moving the quadrupoles instead of the screen, we
are varying the initial Twiss parameters. However, since we
know the initial  function of the input beam to be at least
1 m, and since the scans must only cover several 
, or less
than 1 cm, the initial conditions remain essentially constant
over the scan.

072401-15
J. K. LIM et al.
Thus, the analysis of the moving screen technique to the
‘‘moving quads’’ measurement is valid in the present sce-
nario. One such dataset and analysis is shown in Fig. 22.
The effects of limited image resolution can be clearly seen
in these measurements. The expected parabolic depen-
dence of 2 on z is observed well away from the minimum
beam sizes, but the data give a nearly flat dependence in the
zone near z0 . In the case studied in Fig. 22, normalized
initial emittances measured at 58.2 MeV beam energy
using the EM quad triplet are "x  16:6 mm mrad and
"y  9:7 mm mrad, while the normalized emittances ob-
tained with the PMQ scan are "x  16:3 mm mrad and
"y  12:9 mm mrad. During the initial period of running
the final-focus system, beam spot sizes as low as rms  pump-probe experiments which investigate dynamic dif-
18 m were measured in both transverse dimensions after
optimization.
As noted, this is an upper limit on the actual beam size,
due to the resolution limitations in the measurement sys-
tem. We note from Fig. 22 that the value of the res
obtained from the best fit of the quad scan data to
Eq. (14) is also 18 m. With the value of res determined
by the fit to the data, we deduce that the minimum beam
size in each transverse dimension in Fig. 22 would have
been in the range of 15 m in the absence of image
resolution problems (e.g., geometric and chromatic aberra-
tions). In order to observe such small sizes, a reconfiguring
of the optics between the interaction region OTR surface
and the viewing CCD camera will be undertaken in the
future. One of the most critical needed capabilities the
PMQ-based final-focus system, that of tunability over a
wide range of beam energies, has been demonstrated.
-9
5x10
=12.9 (mm mrad),
n,y
-9
4x10 =1.79 mm,
y x
=14.2 m
min
-9
3x10
(m )
2
2
-9
2x10
-9
1x10
0
0 10
-0.002 0 0.002
FIG. 22. (Color) Quadrupole scan of the final-focus beam at LLNL PLEIADES, performed by moving the entire PMQ assembly.
072401-16
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
Through longitudinal motion of the PMQs, the final-focus
system was capable of adjusting to consistently produce a
small beam spot size at different input beam energies. This
variability, over 40% in beam energy, has produced varia-
tion of x-ray photon energies over the range between 40 to
140 keV.
Recently, the final-focus system was successfully recon-
figured by boring out the initial isolated PMQ’s inner
diameter to lower its gradient, and removing one of each
of the downstream PMQ pairs. This new system has al-
lowed production of an optimum beam focus at as low as
20 MeV electron beam energy. This new capability has
allowed use of PLEIADES to pursue ongoing dynamic
fraction in different heavy-metal materials.
VII. DISCUSSIONS AND CONCLUSIONS
An adjustable short focal length final-focus system
based on an asymmetric (F-DD-FF) triplet configuration
composed of unprecedented ultrahigh field gradient, with
experimental value 560 T=m, PMQs has been developed.
The design issues concerning the PMQs themselves were
studied in detail using the 3D magnetostatic simulation
code RADIA. The PMQ array was fabricated, measured, and
placed on a precision moving system. It was then imple-
mented at the ICS final-focus system beam line at LLNL
PLEIADES. This effort led to creation of ultrasmall beam
spot sizes, and an enhanced performance of a high-
brightness ICS x-ray source, with adjustability demon-
strated over a wide range of input beam energies, and
thus produced x-ray energies.
=16.3 (mm mrad),
n,x
=1.62 mm,
=15.2 m
min
x
y
0.004 0.006 0.008 0.01 0.012
z (m)
ADJUSTABLE, SHORT FOCAL LENGTH PERMANENT- . . .
The principles behind the short focal length, adjustable
PMQ-based final-focus system are quite attractive for ad-
vanced applications. At UCLA, a similar system, with few
cm focal lengths, is presently being implemented in the
final focus of a nonlinear ICS experiment in the Neptune
laboratory. As the emittance is expected to be a factor of 3
lower in the Neptune experiment, this test should allow
demonstration of the emittance preservation at a level
much closer to the state-of-the-art in high-brightness pho-
toinjector systems. Additionally, the current experience in
creating very high gradient, small-bore quadrupoles is of
continuing interest for application to linear collider final-
focus systems, as the demonstrated gradient achieved here
exceeded the CLIC efforts by over 15%. Finally, we note
that standard F-DD-F PMQ triplets using longitudinal
position tuning, with much larger bore and lower gradients
have been used at LANSCE for proton radiographic mi-
croscopy [26], an application which does not require large
demagnification, but nonetheless demands very high
strength quadrupoles.
ACKNOWLEDGMENTS
The authors would like to thank G. Andonian and J.
England for their assistance with simulations. The authors
also thank G. Anderson for the assistance in fielding ex-
periments. This work was performed under the auspices of
the U.S. Department of Energy under Contracts No. DE-
FG-98ER45693 and No. DE-FG03-92ER40693, and by
Lawrence Livermore National Laboratory under Contract
No. W-7405-Eng-48 and the LLNL ILSA program under
Contract No. LS04-001-B.
[1] J. S. Fraser, R. L. Sheffield, E. R. Gray, and G. Rodenz, in
IEEE Trans. Nucl. Sci. 32, 1791 (1985).
[2] P. Chen, J. M. Dawson, R. W. Huff, and T. Katsouleas,
Phys. Rev. Lett. 54, 693 (1985).
[3] J. B. Rosenzweig, B. Breizman, T. Katsouleas, and J. J. Su,
Phys. Rev. A 44, R6189 (1991).
[4] R. W. Schoenlein et al., Science 274, 236 (1996).
[5] I. V. Pogorelsky et al., Phys. Rev. ST Accel. Beams 3,
090702 (2000).
[6] L. Serafini and J. B. Rosenzweig, Phys. Rev. E 55, 7565
(1997).
072401-17
Phys. Rev. ST Accel. Beams 8, 072401 (2005)
[7] J. B. Rosenzweig and P. Chen, Phys. Rev. D 39, 2039
(1989).
[8] S. Anderson et al., Appl. Phys. B 78, 891 (2004).
[9] D. J. Gibson et al., Phys. Plasmas 11, 2857 (2004).
[10] W. Lou, D. Hartill, D. Rice, D. Rubin, and J. Welch, in
Proceedings of the Particle Accelerator Conference,
Vancouver, BC, Canada, 1997 (IEEE, Piscataway, NJ,
1998), Vol. 3, p. 3236.
[11] J. T. Volk et al., in Proceedings of the Particle Accelerator
Conference, Chicago, IL, 2001 (IEEE, Piscataway, NJ,
2001), Vol. 1, p. 217.
[12] F. Zimmermann, R. Assmann, G. Guignard, D. Schulte,
and O. Napoly, in Proceedings of the European Particle
Accelerator Conference, Vienna, 2000 (EPS, Geneva,
2000).
[13] F. Zimmermann, R. Assmann, G. Guignard, D. Schulte,
and O. Napoly, CERN Technical Report No. CERN-SL-
2000-057 AP, 2000.
[14] S. C. Gottschalk, W. D. Kimura, and D. H. Dowell, in
Proceedings of the Particle Accelerator Conference,
Chicago, IL, 2001 (Ref. [11]), Vol. 5, p. 3218.
[15] R. B. Brown and J. R. Cost, IEEE Trans. Magn. 25, 3117
(1989).
[16] K. Halbach, Nucl. Instrum. Methods 169, 1 (1980).
[17] W. Lou, D. Hartill, D. Rice, D. Rubin, and J. Welch, Phys.
Rev. ST Accel. Beams 1, 022401 (1998).
[18] P. Elleaume, O. Chubar, and J. Chavanne, in Proceedings
of the Particle Accelerator Conference, Vancouver, BC,
Canada, 1997 (Ref. [10]), Vol. 3, p. 3509.
[19] J. Chavanne, O. Chubar, P. Elleaume, and P. V.
Vaerenbergh, in Proceedings of the European Particle
Accelerator Conference, Vienna, 2000 (Ref. [12]), p. 2316.
[20] K. Crandall and D. Rusthoi, Los Alamos National
Laboratory Technical Report No. LA-UR-97-886, 1997.
[21] M. Borland, Advanced Photon Source Technical Report
No. LS-287, 2000.
[22] A. Chao and M. Tigner, Handbook of Acccelerator
Physics and Engineering (World Scientific, Singapore,
1999).
[23] R. Warren, Nucl. Instrum. Methods Phys. Res., Sect. A
272, 257 (1988).
[24] D. T. Palmer, S. Anderson, and J. B. Rosenzweig, in
Physics of High Brightness Beams (World Scientific,
Singapore, 2000), p. 439.
[25] J. B. Rosenzweig, Fundamentals of Beam Physics (Oxford
University Press, Inc., New York, 2003).
[26] T. Mottershead et al., in Proceedings of the Particle
Accelerator Conference, Portland, OR, 2003 (IEEE,
Piscataway, NJ, 2003), Vol. 1, p. 702.